Most schools do run a low pass filter on your GRE scores (if required) and GPA. The GPA is weighted according the repuation of your school. However, once you get beyond the filtering, these things don't matter.

The most important thing at a top PhD school is letters of recommendations from faculty, with the greatest weight given to letters from a professor they might know. ("They" being any of the members of the admissions committee who happen to read your file, or are present during the discussion. There is a fair amount of luck involved). This maybe hard for you to get if the professors at your current university do not publish (i.e., attend conferences, etc.). However, no matter what, do keep in touch with them. You'll need at least 2-3 letters from them. Letters from work will be fine, but do not carry heavy weight -- especially beyond one letter.

The other relevant things include: the school you attended, research experience (publications are your greatest asset), and lastly, your statement of purpose. I don't think anything else really matters too much -- at least to a top CS program.

But if you really want to go to grad school, there is always a way. Some tips:

(1) Apply to a lot of schools, as the admissions process can be fairly random. However, do not apply to safe schools for the sake of applying.

(2) Also, an alternative is to apply for a Masters program. Get in. Impress a professor or two. Upgrade to PhD program. Obviously, it is hard to find financial support for Masters student at some schools. If you are local to a great university, you can also take graduate courses a non-degree student (very easy to get in, as schools love money), impress one or two professors, and apply.

PS: I served as a student representative on an admissions committee at UIUC in the past. I also came from a "lowly" state university. However, I jumped to PhD program directly after my BS.

Feel free to shoot me an email. My email is in my profile.

Also, note that the top-tier programs generally don't have academic Master's programs. Stanford offers one but it's career-oriented.

You could grab a research position with the government. Lincoln Labs at Hanscom AFB is associated with MIT for example.

If you work a few years as a program manager or bench scientist in the government civil service, you will network with prominent folks and have a chance to gain their respect. After all--where do you think that research money that professors love so much comes from?

I had a positive experience. I was able to network with people who wrote papers I really liked. One particularly touching experience--months after I left, someone contacted me to let me know that a Professor whose work influenced me quite a bit would be in town. I was (as an outsider) given a slice of his schedule to meet with him and have lunch. Its like a family, once you're in... you're in.

You should probably ask your current professors for letters of recommendation now and have them sent to your registrar so they can be forwarded on later when you decide to apply. Unless you have a natural way to continue your connection with your professors it will likely be very difficult to go back to them even a year later and ask for the letter and have them remember you (unless you were a super-star of course). Just get your letters done now. You can always have them update the letter in x years if you feel there's something relevant they can add. More likely than not you'll just be thankful that you already have the letters since you never really maintained contact. On the other hand, remember that the longer you wait, the more unusual it's going to be for the admission committee to see a letter from so many years back.

If you're really serious about wanting to get into one of those programs you could actually move to Stanford/MIT and try to get a research position working with some of the people in those departments. Or perhaps going to them and asking if there's any companies they would recommend working at prior to a graduate program.

Also, I have to put a disclaimer here that getting a PhD will probably not be worth it from a financial standpoint, and most likely you will not find a tenure track position afterwards.

0. If you want answers as to what schools look for in students start by looking at CV's of students in the department. More specifically, look at students in the research areas you may be interested in. This is probably the single best resource available to you. I knew my weaknesses (non top 10 undergraduate school, several years in industry, no undergraduate research experience, relatively unknown letter writers, and a non-CS undergraduate degree) and adjusted my expectations accordingly (ie I knew I wasn't going to get in a top 10 school despite high grades, good industry job, and perfect quant gre).

1. 0 years. There are a number of reasons from familiarity with your letter writers to the lack of commitment if you work in a non-research position.

2. If you are working in a research position and could publish that is ideal. However, I doubt this type of position will be available to you (usually it requires a phd). In terms of acceptance I believe industry experience is nearly meaningless. There are plenty of reasons to do it for personal reasons, but look at the CV's of current students at the schools you are targeting and look at how many of them worked between undergrad and grad.

3. Again, a research position is probably the only type of position that will help you in the admissions process. It may give you access to significant letter writers, a chance to be an author, etc..

4. I presume that status implies a significant contribution. This could help, but unless the contribution is related to computer science it probably isn't worth much.

Good luck.

edit :: a couple small updates.

1. Don't wait. Opportunity costs are low right now (bad economy) and waiting hurts your chances. If you wait at least 5 years, you are > 27 when you start grad school. You graduate at age 32. That's over the hill. Many grad schools will flatly reject you for this reason.

2. Publications might help, if they are solid technical works in peer reviewed journals. Industry experience could also help IF it's hardcore R&D work. Academics care relatively little about open source, unless you did something truly awesome (e.g., FFTW).

3. Yes. R&D type jobs are the only jobs that won't hurt you when applying to grad school.

4. GvR could probably get into a good grad school. Below that will probably not help very much.

However, many places will give you a masters if you pay tuition. Don't expect to jump from the masters to Ph.D. track, however.

Regarding legality, I actually made a mistake when I said reject due to age. I meant reject due to having stale ideas and not appearing committed enough, or just not standing out from the crowd.

(1) Admissions process at many places: committee weeds out the weakest and professors/groups pick their students from the remainder. If a professor in your field doesn't like 30 year olds, you failed to stand out from the crowd.

I hope this isn't universal. I am planning to apply in 2010 and I'll be 38 when I do. Oh well one more wall to jump over, so what's new? :-)

Of course, it could be an exception, but I'm sure admission was granted without respect to age. As far as I know though, luck was a big determining factor, and as such, I would look at a lot of these things (particularly if you aren't a monstrously strong candidate) very lightly.

Hey Thanks!

"particularly if you aren't a monstrously strong candidate"

Good Point.

A "monstrously strong" candidate is what I am trying to be. Hey if we ask startup founders to be monstrously strong developers why not hold ourselves to the same standards as grad students. The journey is very interesting, forcing me to evaluate my deficiencies as a candidate and get better constantly.

A scientist I am working with on some research software said recently, "You'd make a great PhD student at Carnegie Mellon" (where he got his PhD) and has promised to write a reccomendation, so there's some progress. We'll see how it goes.

Mitch Resnick, professor at MIT Media Lab, finished his PhD when he was 36. Fourteen years after his undergraduate degree, meanwhile spending five years as a science/technology journalist for Business Week magazine.

And he has done some great research even while being older than a typical graduate student.

http://en.wikipedia.org/wiki/Mitchel_Resnick

but otherwise it'll be a long time until you're able to payback the lost opportunity costs of being in grad school (no career advancement, little pay), since other than being a professor, it's unlikely having a PhD will get you any different job than before (if you're in CS).

(I was a PhD student at a top 3 cs program and left for a startup years ago)

That is exactly true for me, except obviously the "young" part :-). (But hey i feel young!)

I want to do a PhD for the experience of doing/learning to do research with top notch people and (corny as it sounds) advancing the frontiers of human knowledge. I don't really care about becoming a prof.

As for "productive", I've made enough money for me to live on comfortably, if not extravagantly , for the next twenty years. (India is a relatively cheap place to live if you have your own house and a good chunk of money saved). I am single and plan to stay that way so that obviously helps.

Research is just something I want to do (very well) before I die.

Will this person be a good academic researcher? To figure this out, they will look at:

1. Can you already do research? If you've published at academic research conferences, they'll read your paper and judge you by it. If you have good recommendations from people they trust, or are known in their field, they will read those and rely on them heavily. Otherwise they view your application as a crapshoot (why would they want to take on a student they're not sure will succeed?)

2. grades, programming abilities, and other things are all secondary. If they pass the bar, that's good, but they're not going to get you in.

My advice: get a programming job working in a research lab, then apply to places the people you're working with have worked with before. Look at MSR (Microsoft lABS), Intel Research, and Google. Otherwise go to a startup.

If you are doing things that are not academia oriented, the longer you do those things, the less chance you will have at acceptance.

"I'm obviously going to continue the work that I enjoy doing."

Careful, academia may not be right for you :-)

"Publications, open source contributions, industry experience?"

Publications will by far trump anything else for most departments. And letters of recommendation (preferably from people the department has heard of...).

"Should I take the time to find a research based position at a company like IBM or HP?"

This is probably the only thing that would realistically help you. You would need to stay at one place long enough to contribute directly to refereed papers, this will really help your application. There are a lot of research positions at places that are not big companies, too, and those tend to submit more papers. Often they are looking for staff programmers and you can "get in on" some papers over time as you work hard and make good observations etc. Or you could be the assistant who is writing the performance harness, etc.

The hard part would be getting in the door probably, in my experience anything CS related will usually accept CS graduate students into internships etc. but finding a full time job there with a BS and no previous CS research experience will be tougher.

---

Academic departments typically only care that you can code well enough to pass their classes. You need to be able to program, for sure.

The better you are, the easier it will be to get through school -- you will need all the time you can get in order to concentrate on algorithms, linear algebra, etc. But it probably won't you help you too much at application time (unless maybe you have won programming competitions).

"Does status in the software community affect acceptance into a CS program?"

Probably not too much, unless the program is linked to a software product (like where I work, the University of Chicago CS department and the Globus grid computing community are intimately tied).

It's been nearly seven years for me now. I've stayed only slightly in touch with one of my undergraduate professors. I feel that getting academic recommendation letters may be difficult at this point. If you're going to wait, I would suggest maintaining conversation with the professors most able to write letters for you.

1. PUBLISH! 2. PUBLISH!! 3. PUBLISH!!!

When I'm reviewing applicants for admission, I look at three things: (1) prior research, (2) letters of recommendation and (3) personal statement. I look at them in that order.

I don't care whether you came from MIT or Bumblefark State U. I don't care whether your GPA was a 2.0 or a 4.0. I don't care whether your Math GRE is an 800 or a 600. I won't look at your transcript either.

I care only about your potential to do research with me. (Keep this in mind when you're choosing which areas and which professors you'd like to work with on the application; it will determine who sees your application.)

My last piece of advice would be to consider a broad range of schools. The rankings put out by US News are completely off base. Many "lowly ranked" schools have deep strengths in particular areas.

Northeastern University, for instance, has an outstanding programming languages faculty, but they get no credit for it in the rankings.

I'll also plug the University of Utah as a great place to come for programming languages/compilers, formal methods, graphics and scientific computing.

Ho ho ho. After the ex-Bear and Lehman experienced coders find jobs, then maybe the finance industry will start hiring entry-level coders again. Maybe.

Ask yourself tho' what you want from a PhD. It's not a "higher qualification" in the sense that a MSc is, learning specialist knowledge to do a niche job. It's an apprenticeship to become an academic. Where do you see yourself post-PhD? What is the outcome you are looking to achieve here? Altho' grad students may write an awful lot of code, no-one really cares about it - its only purpose is to demonstrate theoretical ideas. Coding experience gained in industry won't count for much. Not even Open Source coding. Computer Science is as related to software development as Astronomy is to telescope manufacturing. Working in a Carl Zeiss factory on the shop floor wouldn't count for much applying for an Astronomy PhD would it?

Getting a PhD involves very focused research. You want to make sure that you will truly love the field before you go into it.

I strongly recommend that you take an R&D job at a well-known company if you can. These large companies often sponsor research with top-notch universities. You might be able to use these relationships to get in. My company's sponsored research with Stanford allows me to work with a professor and grad student there. That can't hurt during the application process.

Also, many of these large, research oriented companies will pay for graduate education, although that will probably drop off as the economy tanks.

The only downside to getting a job at a large, research oriented company is that these places are often cube farms that epitomize office space and dilbert comic strips. I guess thats why you leave after a few years and go to grad school

2. You probably won't be publishing when you're working, and open-source projects are good but not enough to put you over the cusp.

3. If you can get a research position, that'll be better for your grad school prospects than finance.

4. I don't know the answer to this one.

Caveat: I was in a math PhD program for a year, and CS may be different.